>
> Humm, did you test it (by unplugging the sda disk and booting) ?
>
Yes, worked just fine. The raid code figured out that the device (now
/dev/sda3) used to be /dev/sdb3, updated the superblock accordingly, and
continued in degraded mode.
> If you didn't you might be in for a surprise...
>
> If you didn't mount /dev/sdb1 as /boot before running lilo to put a MBR
> into /dev/sdb, lilo will write a MBR in /dev/sdb witha first stage
> loader that will look in /dev/sda1 for the second stage boot loader, map
> and kernel at boot time.
I didn't need to do that because /dev/sdb1 is an exact copy of /dev/sda1.
Yes, the second stage boot loader will see the boot partition as /dev/sda1,
but that's just what I want anyway. If I do have to reboot after a failed
first disk (/dev/sda), the second disk will become the first disk
(/dev/sdb --> /dev/sda). So it works because /dev/sdb1 is an exact
duplicate of /dev/sda1.
>
> I've never had any success putting MBRs in /dev/sbd while /dev/sda was
> running. If anybody actually have done it, I would like to know how.
It worked for me.
I'm not saying I won't need a boot disk, just not for this particular
failure scenario.
Regards,
Mike
--
Michael Milligan - Acme Byte & Wire LLC - [EMAIL PROTECTED]
